Auditory brainstem response (ABR) is an objective method for detecting hearing loss, which is widely used in clinical practice, and its waveform characteristics are influenced by stimulus parameters. There are stimulus artifacts in ABR measured using unipolar stimulus, and alternating polarity is currently the only way to eliminate stimulus artifacts. However, considering the physiological differences in the effects of stimuli with different polarities on the auditory system, alternating polarity stimuli may lead to latency jitter in the induced ABR. Therefore, a new method was proposed to eliminate stimulus artifacts - the method of division and sum polarity, which first used positive and negative stimuli separately and then superimposed the two responses induced. The subjects with normal hearing were recruited, and their data that the click ABRs under four polarity ways (positive polarity, negative polarity, alternating polarity, division and sum polarity), as well as the tone-burst ABRs at five frequencies under two polarity ways (alternating polarity and division and sum polarity) were compared, with a focus on their waveform differentiation and latency differences. The results show that the difference in click ABRs under different polarity ways is insignificant, indicating that it is not sensitive to stimulus polarity. The waveform of low-frequency tone-burst ABRs is better under division and sum polarity than under alternating polarity, indicating that the low-frequency tone-burst ABR is more sensitive to stimulus polarity, and the sensitivity decreases with the increase of stimulus frequency. Based on the analysis of the above results, it is recommended to use unipolar stimulus for the click ABR, and the division and sum polarity method for the tone-burst ABR. The feasibility of the division and sum polarity method is validated in this study, which provides a new approach for eliminating stimulus artifacts when measuring evoked potentials.